1656 lines
28 KiB
Markdown
1656 lines
28 KiB
Markdown
|
|
# Kairo – Universal Steering Engine Decision
|
|||
|
|
## v0.1 – Fachliche Entscheidung zu Steuerungsengine, Roadmap, Methoden, Workflow-Nodes und Agenten-Einbindung
|
|||
|
|
|
|||
|
|
**Status:** Entscheidungs- und Spezifikationsentwurf
|
|||
|
|
**Stand:** 2026-07-05
|
|||
|
|
**Kontext:** Nach AP0.7, nach Product Reset, vor Überarbeitung von AP0.8
|
|||
|
|
**Ziel:** Verbindliche fachliche Grundlage für die spätere universelle Steuerungsengine von Kairo, ohne AP0.8 technisch zu überladen.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 1. Zweck dieses Dokuments
|
|||
|
|
|
|||
|
|
Dieses Dokument verdichtet die bisherige fachliche Diskussion zu einer entscheidungsfähigen Spezifikation.
|
|||
|
|
|
|||
|
|
Es beantwortet insbesondere:
|
|||
|
|
|
|||
|
|
1. Welche Verantwortung hat Kairo selbst?
|
|||
|
|
2. Welche Verantwortung kann an externe Agentensysteme ausgelagert werden?
|
|||
|
|
3. Wie werden unterschiedliche Entwicklungs- und Umsetzungsmethoden fachlich modelliert?
|
|||
|
|
4. Wie hängen Steering Domains, Lifecycle Models, Methoden, Profile, Roadmaps, Hooks, Workflows, Nodes und Signals zusammen?
|
|||
|
|
5. Wie bleibt Kairo universell erweiterbar, ohne sofort zu komplex zu werden?
|
|||
|
|
6. Was muss AP0.8 vorbereiten?
|
|||
|
|
7. Was darf AP0.8 ausdrücklich noch nicht bauen?
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 2. Ausgangslage
|
|||
|
|
|
|||
|
|
AP0.7 hat wesentliche technische Grundlagen geschaffen:
|
|||
|
|
|
|||
|
|
- Tenant Hardening
|
|||
|
|
- Tenant-Invarianten
|
|||
|
|
- Cross-Tenant-Tests
|
|||
|
|
- Actor Directory
|
|||
|
|
- echter ActorSelect
|
|||
|
|
- Data Layer Minimum
|
|||
|
|
- Workspace Summary
|
|||
|
|
- zentrale Workspace-Read-Endpunkte
|
|||
|
|
- AP0.6b UX/PWA bleibt erhalten
|
|||
|
|
|
|||
|
|
Gleichzeitig ist fachlich klar geworden:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Initiative / Vorhaben
|
|||
|
|
→ Action / Maßnahme
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
ist als technischer Startpunkt akzeptabel, aber als Produktmodell für Kairo zu flach.
|
|||
|
|
|
|||
|
|
Kairo soll kein Aufgabenmanager sein, sondern ein operativer Program Director und langfristig eine universell erweiterbare Steuerungsplattform für Entwicklungen.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 3. Leitentscheidung
|
|||
|
|
|
|||
|
|
Kairo baut langfristig eine **fachliche Steuerungsengine**.
|
|||
|
|
|
|||
|
|
Diese Engine muss unterschiedliche Entwicklungs- und Umsetzungskontexte tragen können:
|
|||
|
|
|
|||
|
|
- Software- und Produktentwicklung
|
|||
|
|
- Buchentwicklung
|
|||
|
|
- Programmarbeit
|
|||
|
|
- Projektarbeit
|
|||
|
|
- persönliche Entwicklung
|
|||
|
|
- Reifegradentwicklung
|
|||
|
|
- Lernpfade
|
|||
|
|
- Routinen
|
|||
|
|
- Reviews / Audits
|
|||
|
|
- Betrieb / Wartung
|
|||
|
|
- Agentenläufe
|
|||
|
|
|
|||
|
|
Die Engine darf jedoch nicht sofort als beliebige freie Workflow-Plattform gebaut werden.
|
|||
|
|
|
|||
|
|
Leitentscheidung:
|
|||
|
|
|
|||
|
|
> Kairo bleibt System of Record und Steering Authority.
|
|||
|
|
> Externe oder integrierte Agentensysteme können ausführende Step Provider sein.
|
|||
|
|
> Die fachliche Steuerung, der Kontext, die Signale, die Roadmap, die Nachvollziehbarkeit und die Verantwortungslogik bleiben in Kairo.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 4. Was Kairo selbst verantwortet
|
|||
|
|
|
|||
|
|
Kairo verantwortet dauerhaft:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- steuerbare Objekte und ihre Beziehungen
|
|||
|
|
- Roadmaps und Entwicklungsziele
|
|||
|
|
- Backlog, Actions, Blocker, Reviews, Evidence und Decisions
|
|||
|
|
- Steering Contexts
|
|||
|
|
- Method Definitions und Method Profiles
|
|||
|
|
- Lifecycle Models und Hook Points
|
|||
|
|
- Signale wie Attention, NextAction, EvidenceRequired, DecisionRequired
|
|||
|
|
- Auditierbarkeit und Nachvollziehbarkeit
|
|||
|
|
- Tenant- und Actor-Sicherheit
|
|||
|
|
- Auswahl und Beauftragung von Human-/Agent-/Tool-Steps
|
|||
|
|
- Rückführung von Ergebnissen in das Kairo-Modell
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Kairo muss also immer wissen:
|
|||
|
|
|
|||
|
|
- Was wird gesteuert?
|
|||
|
|
- Warum wird etwas vorgeschlagen?
|
|||
|
|
- Wer oder was ist verantwortlich?
|
|||
|
|
- Welche Methode gilt?
|
|||
|
|
- Welche Roadmap- oder Entwicklungslogik liegt zugrunde?
|
|||
|
|
- Welcher Workflow oder Hook hat ein Signal erzeugt?
|
|||
|
|
- Welches Ergebnis wurde zurückgeführt?
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 5. Was externe Agentensysteme übernehmen dürfen
|
|||
|
|
|
|||
|
|
Externe oder integrierte Agentensysteme dürfen später einzelne Schritte ausführen, z. B.:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Text analysieren
|
|||
|
|
- Vorschläge generieren
|
|||
|
|
- WBS-Vorschläge erzeugen
|
|||
|
|
- Abhängigkeiten identifizieren
|
|||
|
|
- Backlog Items vorschlagen
|
|||
|
|
- NextActionCandidates vorschlagen
|
|||
|
|
- DoD prüfen
|
|||
|
|
- Evidence auswerten
|
|||
|
|
- Dokumente zusammenfassen
|
|||
|
|
- Recherche durchführen
|
|||
|
|
- Tool Calls ausführen
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Sie dürfen aber nicht ohne Kairo-Kontrolle:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- führendes Steuerungssystem werden
|
|||
|
|
- Tenant-/Actor-Kontext umgehen
|
|||
|
|
- Roadmap oder Methode eigenständig überschreiben
|
|||
|
|
- strategische Prioritäten ändern
|
|||
|
|
- Daten ohne Audit verändern
|
|||
|
|
- Rechte oder Capabilities ändern
|
|||
|
|
- produktive Änderungen ohne vorgesehenes Gate ausführen
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Agentensysteme sind aus Sicht von Kairo:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
ausführende Provider
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
nicht:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
führende Steuerungsinstanz
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 6. Komplexitätsbremse
|
|||
|
|
|
|||
|
|
Die Architektur wird in drei Ebenen geschnitten.
|
|||
|
|
|
|||
|
|
### Ebene 1 – Signals und Read Models
|
|||
|
|
|
|||
|
|
Kurzfristig und MVP-nah.
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Attention
|
|||
|
|
- NextActionCandidate
|
|||
|
|
- EvidenceRequired
|
|||
|
|
- DecisionRequired
|
|||
|
|
- BlockerSignal
|
|||
|
|
- ReviewDue
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Diese Ebene liefert Produktnutzen ohne vollständige Workflow Runtime.
|
|||
|
|
|
|||
|
|
### Ebene 2 – Hook-basierte Workflows
|
|||
|
|
|
|||
|
|
Mittelfristig.
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- feste Hook Points
|
|||
|
|
- definierte Node-Typen
|
|||
|
|
- Workflow Definitions
|
|||
|
|
- Workflow Bindings
|
|||
|
|
- einfache Workflow Runtime
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Diese Ebene erlaubt methodenspezifische Abläufe, aber noch keine beliebige offene Plattform.
|
|||
|
|
|
|||
|
|
### Ebene 3 – konfigurierbare Methoden und Agenten
|
|||
|
|
|
|||
|
|
Langfristig.
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Method Profiles
|
|||
|
|
- tenant-konfigurierbare Regeln
|
|||
|
|
- LLM-/Tool-/Agent-Steps
|
|||
|
|
- externe Agentensysteme
|
|||
|
|
- Method Designer
|
|||
|
|
- Workflow Designer
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
AP0.8 darf maximal Ebene 1 aktiv implementieren und Ebene 2/3 fachlich vorbereiten.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 7. Kernmodell der universellen Steuerung
|
|||
|
|
|
|||
|
|
Das fachliche Zielmodell lautet:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Steerable Object
|
|||
|
|
→ Steering Context
|
|||
|
|
→ Steering Domain
|
|||
|
|
→ Lifecycle Model
|
|||
|
|
→ Steering Method
|
|||
|
|
→ Method Profile
|
|||
|
|
→ Hook Points
|
|||
|
|
→ Rules / Workflow Bindings
|
|||
|
|
→ Workflow Nodes
|
|||
|
|
→ Signals / Results
|
|||
|
|
→ Roadmap / Backlog / Action / Evidence / Review
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Dieses Modell ist bewusst mehrschichtig.
|
|||
|
|
|
|||
|
|
Es verhindert, dass alles auf eine einzige Projektlogik oder eine einzige Aufgabenlogik reduziert wird.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 8. Steerable Object
|
|||
|
|
|
|||
|
|
Ein Steerable Object ist jedes Objekt, für das Kairo Steuerungslogik anwenden kann.
|
|||
|
|
|
|||
|
|
Mögliche Scope Types:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
program
|
|||
|
|
initiative
|
|||
|
|
project
|
|||
|
|
roadmap
|
|||
|
|
roadmap_item
|
|||
|
|
backlog_item
|
|||
|
|
action
|
|||
|
|
blocker
|
|||
|
|
review
|
|||
|
|
evidence
|
|||
|
|
decision
|
|||
|
|
recurring_element
|
|||
|
|
work_cycle
|
|||
|
|
agent_run
|
|||
|
|
custom
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Nicht alle Scope Types müssen in AP0.8 aktiv unterstützt werden.
|
|||
|
|
|
|||
|
|
AP0.8 sollte mindestens vorbereiten:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
initiative
|
|||
|
|
roadmap_item
|
|||
|
|
backlog_item
|
|||
|
|
action
|
|||
|
|
blocker
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 9. Steering Domain
|
|||
|
|
|
|||
|
|
Eine Steering Domain beschreibt die fachliche Domäne der Steuerung.
|
|||
|
|
|
|||
|
|
Initiale Domains:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
program
|
|||
|
|
project
|
|||
|
|
product_development
|
|||
|
|
content_development
|
|||
|
|
personal_development
|
|||
|
|
maturity_development
|
|||
|
|
routine_operations
|
|||
|
|
audit_review
|
|||
|
|
agent_execution
|
|||
|
|
custom
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 9.1 Beispiele
|
|||
|
|
|
|||
|
|
#### Software-/Produktentwicklung
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Domain: product_development
|
|||
|
|
typische Roadmap-Lanes:
|
|||
|
|
- Feature Landscape
|
|||
|
|
- Architecture Evolution
|
|||
|
|
- Releases
|
|||
|
|
- Technical Debt
|
|||
|
|
- Validation
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
#### Buchentwicklung
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Domain: content_development
|
|||
|
|
typische Roadmap-Lanes:
|
|||
|
|
- Buchstruktur
|
|||
|
|
- Kapitel
|
|||
|
|
- Recherche
|
|||
|
|
- Drafts
|
|||
|
|
- Review
|
|||
|
|
- Veröffentlichung
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
#### Persönliche Fähigkeitsentwicklung
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Domain: maturity_development
|
|||
|
|
typische Roadmap-Lanes:
|
|||
|
|
- Reifegradstufen
|
|||
|
|
- Fähigkeiten
|
|||
|
|
- Praxis / Übungen
|
|||
|
|
- Evidence
|
|||
|
|
- Reflexion
|
|||
|
|
- Assessments
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
#### Programmsteuerung
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Domain: program
|
|||
|
|
typische Roadmap-Lanes:
|
|||
|
|
- strategische Ziele
|
|||
|
|
- Initiativen
|
|||
|
|
- Abhängigkeiten
|
|||
|
|
- Ressourcen
|
|||
|
|
- Risiken
|
|||
|
|
- Benefits
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 10. Lifecycle Model
|
|||
|
|
|
|||
|
|
Ein Lifecycle Model beschreibt die typischen Phasen einer Domain.
|
|||
|
|
|
|||
|
|
### 10.1 Project / Initiative Lifecycle
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
intake
|
|||
|
|
clarify
|
|||
|
|
structure
|
|||
|
|
plan
|
|||
|
|
execute
|
|||
|
|
monitor
|
|||
|
|
review
|
|||
|
|
adapt
|
|||
|
|
close
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 10.2 Program Lifecycle
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
define_intent
|
|||
|
|
establish_structure
|
|||
|
|
prioritize_initiatives
|
|||
|
|
coordinate_dependencies
|
|||
|
|
allocate_capacity
|
|||
|
|
monitor_benefits
|
|||
|
|
resolve_conflicts
|
|||
|
|
review_outcomes
|
|||
|
|
rebalance
|
|||
|
|
close
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 10.3 Maturity Development Lifecycle
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
assess_current_level
|
|||
|
|
define_target_maturity
|
|||
|
|
identify_gaps
|
|||
|
|
select_intervention
|
|||
|
|
practice
|
|||
|
|
collect_evidence
|
|||
|
|
reflect
|
|||
|
|
adjust_challenge
|
|||
|
|
reassess
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 10.4 Content Development Lifecycle
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
define_concept
|
|||
|
|
structure_content
|
|||
|
|
research
|
|||
|
|
draft
|
|||
|
|
review
|
|||
|
|
revise
|
|||
|
|
prepare_publication
|
|||
|
|
publish
|
|||
|
|
maintain
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 10.5 Routine Operations Lifecycle
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
define_routine
|
|||
|
|
schedule_next_run
|
|||
|
|
execute_run
|
|||
|
|
capture_exception
|
|||
|
|
verify_completion
|
|||
|
|
review_quality
|
|||
|
|
adjust_routine
|
|||
|
|
repeat
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
AP0.8 muss diese Lifecycles noch nicht technisch als vollständige State Machines implementieren.
|
|||
|
|
|
|||
|
|
Aber die Begriffe und Hook Points müssen fachlich vorbereitet sein.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 11. Steering Context
|
|||
|
|
|
|||
|
|
Ein Steering Context verbindet ein Steerable Object mit Domain, Lifecycle und Methode.
|
|||
|
|
|
|||
|
|
Konzeptionell:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Steering Context
|
|||
|
|
- tenant_id
|
|||
|
|
- scope_type
|
|||
|
|
- scope_id
|
|||
|
|
- steering_domain
|
|||
|
|
- lifecycle_model
|
|||
|
|
- steering_method
|
|||
|
|
- method_profile_id optional
|
|||
|
|
- method_version
|
|||
|
|
- configuration_json
|
|||
|
|
- inherits_from_parent
|
|||
|
|
- is_active
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Mögliche Tabelle:
|
|||
|
|
|
|||
|
|
```sql
|
|||
|
|
steering_contexts (
|
|||
|
|
id UUID PRIMARY KEY,
|
|||
|
|
tenant_id UUID NOT NULL,
|
|||
|
|
|
|||
|
|
scope_type VARCHAR(64) NOT NULL,
|
|||
|
|
scope_id UUID NOT NULL,
|
|||
|
|
|
|||
|
|
steering_domain VARCHAR(64) NOT NULL,
|
|||
|
|
lifecycle_model VARCHAR(64) NOT NULL,
|
|||
|
|
steering_method VARCHAR(128) NOT NULL,
|
|||
|
|
method_profile_id UUID NULL,
|
|||
|
|
method_version VARCHAR(32) NOT NULL DEFAULT 'v1',
|
|||
|
|
|
|||
|
|
configuration_json JSONB NOT NULL DEFAULT '{}'::jsonb,
|
|||
|
|
|
|||
|
|
inherits_from_parent BOOLEAN NOT NULL DEFAULT TRUE,
|
|||
|
|
is_active BOOLEAN NOT NULL DEFAULT TRUE,
|
|||
|
|
|
|||
|
|
created_at TIMESTAMP NOT NULL,
|
|||
|
|
updated_at TIMESTAMP NOT NULL
|
|||
|
|
)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 11.1 Vererbung
|
|||
|
|
|
|||
|
|
Standardregel:
|
|||
|
|
|
|||
|
|
> Untergeordnete Objekte erben den Steering Context ihres Parents, sofern sie keinen eigenen Context besitzen.
|
|||
|
|
|
|||
|
|
Beispiele:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Produktentwicklung Kairo
|
|||
|
|
Initiative: product_development / milestone_driven
|
|||
|
|
RoadmapItem "Feature Landscape": geerbt
|
|||
|
|
BacklogItem "Attention DTO definieren": geerbt
|
|||
|
|
Action "Endpoint bauen": simple task
|
|||
|
|
|
|||
|
|
Buchentwicklung
|
|||
|
|
Initiative: content_development / chapter_based
|
|||
|
|
RoadmapItem "Kapitel 3": content_draft
|
|||
|
|
Action "Rohfassung schreiben": writing_task
|
|||
|
|
|
|||
|
|
Persönliche Entwicklung
|
|||
|
|
Initiative: maturity_development / maturity_progression
|
|||
|
|
RoadmapItem "Reifegrad Beweglichkeit 2": maturity_stage
|
|||
|
|
Action "Dehnroutine durchführen": routine_task
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 12. Steering Method
|
|||
|
|
|
|||
|
|
Eine Steering Method beschreibt, nach welcher Logik gesteuert wird.
|
|||
|
|
|
|||
|
|
Beispiele:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
simple
|
|||
|
|
milestone_driven
|
|||
|
|
kanban
|
|||
|
|
cycle_based
|
|||
|
|
chapter_based
|
|||
|
|
maturity_progression
|
|||
|
|
habit_based
|
|||
|
|
review_driven
|
|||
|
|
audit_driven
|
|||
|
|
recurring_control
|
|||
|
|
agent_controlled
|
|||
|
|
hybrid
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Eine Methode definiert fachlich:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- unterstützte Domains
|
|||
|
|
- unterstützte Scope Types
|
|||
|
|
- typische Roadmap-Lanes
|
|||
|
|
- typische RoadmapItem-Typen
|
|||
|
|
- relevante Hook Points
|
|||
|
|
- erlaubte oder empfohlene Node-Typen
|
|||
|
|
- relevante Signals
|
|||
|
|
- Fortschrittslogik
|
|||
|
|
- Reviewlogik
|
|||
|
|
- Evidence-Anforderungen
|
|||
|
|
- NextAction-Logik
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 13. Method Definition
|
|||
|
|
|
|||
|
|
Eine Method Definition ist die systemische Beschreibung einer Methode.
|
|||
|
|
|
|||
|
|
Kurzfristig kann sie code-first sein.
|
|||
|
|
|
|||
|
|
Langfristig kann sie als DB-Definition oder konfigurierbares Profil existieren.
|
|||
|
|
|
|||
|
|
Mögliche Struktur:
|
|||
|
|
|
|||
|
|
```json
|
|||
|
|
{
|
|||
|
|
"method_key": "maturity_progression",
|
|||
|
|
"version": "v1",
|
|||
|
|
"label": "Reifegradentwicklung",
|
|||
|
|
"supported_domains": ["maturity_development", "personal_development"],
|
|||
|
|
"supported_scope_types": ["initiative", "roadmap_item", "action"],
|
|||
|
|
"default_lifecycle_model": "maturity_lifecycle",
|
|||
|
|
"roadmap_lane_types": ["maturity_stages", "capabilities", "practice", "evidence", "reflection"],
|
|||
|
|
"roadmap_item_types": ["maturity_stage", "capability", "learning_step", "review_gate"],
|
|||
|
|
"hook_points": ["on_evidence_required", "on_review_due", "on_next_action_requested"],
|
|||
|
|
"signal_types": ["attention", "next_action", "evidence_required", "review_due"],
|
|||
|
|
"node_types": ["human_task", "evidence_check", "review", "llm_prompt"],
|
|||
|
|
"maturity": "basic"
|
|||
|
|
}
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 14. Method Profile
|
|||
|
|
|
|||
|
|
Ein Method Profile ist eine konkrete konfigurierte Ausprägung einer Methode.
|
|||
|
|
|
|||
|
|
Beispiele:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
cycle_based:
|
|||
|
|
Zwei-Wochen-Zyklus mit Reviewpflicht
|
|||
|
|
|
|||
|
|
maturity_progression:
|
|||
|
|
Vierstufiges Reifegradmodell mit Evidence und Reflexion
|
|||
|
|
|
|||
|
|
content_development:
|
|||
|
|
Buchentwicklung mit Kapitelstruktur, Recherche, Review und Revision
|
|||
|
|
|
|||
|
|
recurring_control:
|
|||
|
|
Monatliche Wartungsroutine mit Eskalation bei verpasstem Durchlauf
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Method Profiles brauchen langfristig:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- tenant_id optional
|
|||
|
|
- method_key
|
|||
|
|
- version
|
|||
|
|
- name
|
|||
|
|
- configuration_json
|
|||
|
|
- allowed_node_types
|
|||
|
|
- enabled_hooks
|
|||
|
|
- default_workflows
|
|||
|
|
- evidence_policy
|
|||
|
|
- review_policy
|
|||
|
|
- approval_policy
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
AP0.8 muss Method Profiles noch nicht vollständig implementieren.
|
|||
|
|
|
|||
|
|
Aber die Architektur muss sie vorsehen.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 15. Development Roadmap
|
|||
|
|
|
|||
|
|
Jede größere Entwicklung braucht eine grobe Roadmap.
|
|||
|
|
|
|||
|
|
Die Roadmap ist nicht gleich Backlog und nicht gleich Maßnahmenliste.
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
RoadmapItem = Entwicklungsziel / Orientierung / geplanter Entwicklungsschritt
|
|||
|
|
BacklogItem = konkreter möglicher Handlungsbedarf
|
|||
|
|
Action = freigegebene operative Maßnahme
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 15.1 Roadmap-Struktur
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Steerable Object
|
|||
|
|
→ Roadmap
|
|||
|
|
→ Roadmap Lane
|
|||
|
|
→ Roadmap Item
|
|||
|
|
→ BacklogItem
|
|||
|
|
→ Action
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 15.2 Warum Roadmap universell sein muss
|
|||
|
|
|
|||
|
|
Eine Softwareentwicklung braucht z. B.:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Feature Landscape
|
|||
|
|
- Architekturentwicklung
|
|||
|
|
- Releases
|
|||
|
|
- technische Schulden
|
|||
|
|
- Validierung
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Eine Buchentwicklung braucht:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Buchstruktur
|
|||
|
|
- Kapitel
|
|||
|
|
- Recherche
|
|||
|
|
- Drafts
|
|||
|
|
- Review
|
|||
|
|
- Veröffentlichung
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Eine persönliche Entwicklung braucht:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Reifegradstufen
|
|||
|
|
- Fähigkeiten
|
|||
|
|
- Übungs-/Praxislinien
|
|||
|
|
- Evidence
|
|||
|
|
- Reflexion
|
|||
|
|
- Assessments
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Deshalb darf Roadmap nicht nur aus Milestones bestehen.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 16. Roadmap-Datenmodell
|
|||
|
|
|
|||
|
|
### 16.1 roadmaps
|
|||
|
|
|
|||
|
|
```sql
|
|||
|
|
roadmaps (
|
|||
|
|
id UUID PRIMARY KEY,
|
|||
|
|
tenant_id UUID NOT NULL,
|
|||
|
|
scope_type VARCHAR(64) NOT NULL,
|
|||
|
|
scope_id UUID NOT NULL,
|
|||
|
|
title VARCHAR(255) NOT NULL,
|
|||
|
|
roadmap_type VARCHAR(64) NOT NULL,
|
|||
|
|
status VARCHAR(32) NOT NULL,
|
|||
|
|
created_at TIMESTAMP NOT NULL,
|
|||
|
|
updated_at TIMESTAMP NOT NULL
|
|||
|
|
)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
`roadmap_type`:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
product
|
|||
|
|
program
|
|||
|
|
project
|
|||
|
|
content
|
|||
|
|
personal_development
|
|||
|
|
maturity
|
|||
|
|
operations
|
|||
|
|
learning
|
|||
|
|
custom
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 16.2 roadmap_lanes
|
|||
|
|
|
|||
|
|
```sql
|
|||
|
|
roadmap_lanes (
|
|||
|
|
id UUID PRIMARY KEY,
|
|||
|
|
tenant_id UUID NOT NULL,
|
|||
|
|
roadmap_id UUID NOT NULL,
|
|||
|
|
title VARCHAR(255) NOT NULL,
|
|||
|
|
lane_type VARCHAR(64) NOT NULL,
|
|||
|
|
sort_order INTEGER NOT NULL DEFAULT 0,
|
|||
|
|
created_at TIMESTAMP NOT NULL,
|
|||
|
|
updated_at TIMESTAMP NOT NULL
|
|||
|
|
)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
`lane_type`:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
milestones
|
|||
|
|
features
|
|||
|
|
maturity_stages
|
|||
|
|
capabilities
|
|||
|
|
chapters
|
|||
|
|
research
|
|||
|
|
cycles
|
|||
|
|
releases
|
|||
|
|
learning_path
|
|||
|
|
architecture
|
|||
|
|
risks
|
|||
|
|
dependencies
|
|||
|
|
custom
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 16.3 roadmap_items
|
|||
|
|
|
|||
|
|
```sql
|
|||
|
|
roadmap_items (
|
|||
|
|
id UUID PRIMARY KEY,
|
|||
|
|
tenant_id UUID NOT NULL,
|
|||
|
|
roadmap_id UUID NOT NULL,
|
|||
|
|
lane_id UUID NULL,
|
|||
|
|
parent_item_id UUID NULL,
|
|||
|
|
title VARCHAR(255) NOT NULL,
|
|||
|
|
description TEXT DEFAULT '',
|
|||
|
|
item_type VARCHAR(64) NOT NULL,
|
|||
|
|
status VARCHAR(32) NOT NULL,
|
|||
|
|
target_date DATE NULL,
|
|||
|
|
sort_order INTEGER NOT NULL DEFAULT 0,
|
|||
|
|
metadata_json JSONB NOT NULL DEFAULT '{}'::jsonb,
|
|||
|
|
created_at TIMESTAMP NOT NULL,
|
|||
|
|
updated_at TIMESTAMP NOT NULL
|
|||
|
|
)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
`item_type`:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
milestone
|
|||
|
|
maturity_stage
|
|||
|
|
feature
|
|||
|
|
capability
|
|||
|
|
phase
|
|||
|
|
release
|
|||
|
|
chapter
|
|||
|
|
research_topic
|
|||
|
|
work_cycle
|
|||
|
|
learning_step
|
|||
|
|
review_gate
|
|||
|
|
decision_gate
|
|||
|
|
architecture_step
|
|||
|
|
dependency
|
|||
|
|
risk_reduction
|
|||
|
|
custom
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 17. Milestone-Entscheidung
|
|||
|
|
|
|||
|
|
Milestone soll nicht als isoliertes Primärobjekt starten.
|
|||
|
|
|
|||
|
|
Entscheidung:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Milestone = RoadmapItem(type = milestone)
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Begründung:
|
|||
|
|
|
|||
|
|
- Meilensteine sind nur eine RoadmapItem-Art.
|
|||
|
|
- Reifegrade, Features, Kapitel, Releases und Lernschritte brauchen dieselbe Roadmap-Struktur.
|
|||
|
|
- Eine separate Milestone-Tabelle würde später zu Parallelmodellen führen.
|
|||
|
|
- RoadmapItems erlauben universellere Entwicklungspfade.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 18. Backlog und Action
|
|||
|
|
|
|||
|
|
### 18.1 BacklogItem
|
|||
|
|
|
|||
|
|
Ein BacklogItem ist ein möglicher Handlungsbedarf.
|
|||
|
|
|
|||
|
|
Beispiele:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Idee
|
|||
|
|
- Risiko
|
|||
|
|
- Vorschlag
|
|||
|
|
- Research-Frage
|
|||
|
|
- Feature-Slice
|
|||
|
|
- Reifegrad-Lücke
|
|||
|
|
- Kapitelbedarf
|
|||
|
|
- Verbesserungsbedarf
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Ein BacklogItem ist noch keine freigegebene Maßnahme.
|
|||
|
|
|
|||
|
|
### 18.2 Action
|
|||
|
|
|
|||
|
|
Eine Action ist eine freigegebene operative Maßnahme.
|
|||
|
|
|
|||
|
|
Sie ist konkret, statusfähig und zuweisbar.
|
|||
|
|
|
|||
|
|
### 18.3 Beziehungen
|
|||
|
|
|
|||
|
|
BacklogItems und Actions können auf RoadmapItems einzahlen.
|
|||
|
|
|
|||
|
|
Mögliche Felder:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
backlog_items.roadmap_item_id NULL
|
|||
|
|
actions.roadmap_item_id NULL
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Langfristig kann ein generisches Relationsmodell folgen.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 19. Blocker
|
|||
|
|
|
|||
|
|
Ein Blocker ist ein eigenes Steuerungsobjekt.
|
|||
|
|
|
|||
|
|
Er kann sich beziehen auf:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Initiative
|
|||
|
|
- RoadmapItem
|
|||
|
|
- BacklogItem
|
|||
|
|
- Action
|
|||
|
|
- WorkCycle
|
|||
|
|
- Review
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
AP0.8 sollte mindestens ermöglichen:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
blockers.initiative_id
|
|||
|
|
blockers.action_id NULL
|
|||
|
|
blockers.roadmap_item_id NULL
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Ein Blocker ist nicht nur `action.status = blocked`.
|
|||
|
|
|
|||
|
|
`action.status = blocked` bleibt ein kompatibler operativer Status, aber der Blocker als Objekt trägt Kontext, Ursache und Steuerungslogik.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 20. Signals
|
|||
|
|
|
|||
|
|
Signals sind erste sichtbare Ergebnisse der Steuerungsengine.
|
|||
|
|
|
|||
|
|
### 20.1 Signal Types
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
attention
|
|||
|
|
next_action
|
|||
|
|
review_due
|
|||
|
|
evidence_required
|
|||
|
|
escalation
|
|||
|
|
blocker_signal
|
|||
|
|
progress_signal
|
|||
|
|
risk_signal
|
|||
|
|
decision_required
|
|||
|
|
dependency_signal
|
|||
|
|
dod_signal
|
|||
|
|
workflow_trigger
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
AP0.8 soll aktiv liefern:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
attention
|
|||
|
|
next_action
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Weitere Signal Types werden fachlich vorbereitet.
|
|||
|
|
|
|||
|
|
### 20.2 AttentionItem
|
|||
|
|
|
|||
|
|
Ein AttentionItem sagt:
|
|||
|
|
|
|||
|
|
> Dieser Steuerungspunkt braucht Aufmerksamkeit.
|
|||
|
|
|
|||
|
|
Beispiele:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- offene Blocker
|
|||
|
|
- blockierte Actions
|
|||
|
|
- RoadmapItem at_risk
|
|||
|
|
- BacklogItem accepted ohne Action
|
|||
|
|
- RoadmapItem planned ohne BacklogItems
|
|||
|
|
- Initiative ohne nächste Maßnahme
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 20.3 NextActionCandidate
|
|||
|
|
|
|||
|
|
Ein NextActionCandidate sagt:
|
|||
|
|
|
|||
|
|
> Dies ist eine sinnvolle nächste Handlung.
|
|||
|
|
|
|||
|
|
Beispiele:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- BacklogItem in Action umwandeln
|
|||
|
|
- Blocker klären
|
|||
|
|
- RoadmapItem strukturieren
|
|||
|
|
- WBS erstellen
|
|||
|
|
- Evidence sammeln
|
|||
|
|
- Review durchführen
|
|||
|
|
- Dependency analysieren
|
|||
|
|
- DoD definieren
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 21. Workflow Engine – Zielbild
|
|||
|
|
|
|||
|
|
Die spätere Workflow Engine sollte nicht als beliebige freie Automatisierungsplattform starten.
|
|||
|
|
|
|||
|
|
Sie sollte fachlich über Hooks und Nodes strukturiert werden.
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Hook Point
|
|||
|
|
→ Workflow Binding
|
|||
|
|
→ Workflow Definition
|
|||
|
|
→ Workflow Node
|
|||
|
|
→ Signal / State Change / Task / Result
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 22. Hook Points
|
|||
|
|
|
|||
|
|
Hook Points sind fachlich stabile Einhängepunkte.
|
|||
|
|
|
|||
|
|
Initiale Hook-Kategorien:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
intake
|
|||
|
|
structure
|
|||
|
|
planning
|
|||
|
|
execution
|
|||
|
|
monitoring
|
|||
|
|
review
|
|||
|
|
adaptation
|
|||
|
|
closure
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Beispiele:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
on_intake_created
|
|||
|
|
on_scope_clarification_needed
|
|||
|
|
on_structure_required
|
|||
|
|
on_wbs_required
|
|||
|
|
on_plan_commit_required
|
|||
|
|
on_dependency_analysis_required
|
|||
|
|
on_next_action_requested
|
|||
|
|
on_execution_started
|
|||
|
|
on_progress_updated
|
|||
|
|
on_blocker_created
|
|||
|
|
on_status_changed
|
|||
|
|
on_dod_definition_required
|
|||
|
|
on_dod_check_required
|
|||
|
|
on_evidence_required
|
|||
|
|
on_review_due
|
|||
|
|
on_replan_required
|
|||
|
|
on_closure_requested
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
AP0.8 muss Hooks noch nicht als ausführbare Workflows implementieren.
|
|||
|
|
|
|||
|
|
Aber Attention/NextAction sollte bereits `hook_key` oder `reason_code` nutzen, damit die spätere Einhängung möglich bleibt.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 23. Workflow Node Types
|
|||
|
|
|
|||
|
|
Die spätere Workflow Engine braucht unterschiedliche Node-Typen.
|
|||
|
|
|
|||
|
|
Initiale fachliche Node-Taxonomie:
|
|||
|
|
|
|||
|
|
### 23.1 Signal Nodes
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
signal_listener
|
|||
|
|
signal_emitter
|
|||
|
|
attention_generator
|
|||
|
|
next_action_generator
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Zweck:
|
|||
|
|
|
|||
|
|
- Signale wahrnehmen
|
|||
|
|
- Signale erzeugen
|
|||
|
|
- Attention Items erzeugen
|
|||
|
|
- NextActionCandidates erzeugen
|
|||
|
|
|
|||
|
|
### 23.2 Rule Nodes
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
rule_evaluator
|
|||
|
|
condition_check
|
|||
|
|
policy_check
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Zweck:
|
|||
|
|
|
|||
|
|
- Regeln prüfen
|
|||
|
|
- Bedingungen auswerten
|
|||
|
|
- Policies anwenden
|
|||
|
|
|
|||
|
|
### 23.3 Human Nodes
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
human_task
|
|||
|
|
approval
|
|||
|
|
decision_gate
|
|||
|
|
review
|
|||
|
|
reflection
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Zweck:
|
|||
|
|
|
|||
|
|
- menschliche Arbeit
|
|||
|
|
- Freigabe
|
|||
|
|
- Entscheidung
|
|||
|
|
- Review
|
|||
|
|
- Reflexion
|
|||
|
|
|
|||
|
|
### 23.4 AI / Agent Nodes
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
llm_prompt
|
|||
|
|
agent_task
|
|||
|
|
agent_review
|
|||
|
|
agent_plan
|
|||
|
|
agent_analysis
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Zweck:
|
|||
|
|
|
|||
|
|
- LLM-Aufruf
|
|||
|
|
- Agentenauftrag
|
|||
|
|
- Agenten-Review
|
|||
|
|
- Plan-/Analysevorschlag
|
|||
|
|
|
|||
|
|
### 23.5 Tool Nodes
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
tool_call
|
|||
|
|
mcp_call
|
|||
|
|
external_api_call
|
|||
|
|
document_lookup
|
|||
|
|
repository_lookup
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Zweck:
|
|||
|
|
|
|||
|
|
- Werkzeuge ausführen
|
|||
|
|
- externe Systeme anbinden
|
|||
|
|
- Dokumente oder Repos durchsuchen
|
|||
|
|
|
|||
|
|
### 23.6 State Nodes
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
state_transition
|
|||
|
|
status_update
|
|||
|
|
assignment_update
|
|||
|
|
roadmap_update
|
|||
|
|
backlog_update
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Zweck:
|
|||
|
|
|
|||
|
|
- Zustände ändern
|
|||
|
|
- Objekte aktualisieren
|
|||
|
|
- Assignments setzen
|
|||
|
|
- Roadmap/Backlog fortschreiben
|
|||
|
|
|
|||
|
|
### 23.7 Validation Nodes
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
evidence_check
|
|||
|
|
dod_check
|
|||
|
|
dependency_analysis
|
|||
|
|
risk_check
|
|||
|
|
quality_gate
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Zweck:
|
|||
|
|
|
|||
|
|
- Nachweise prüfen
|
|||
|
|
- Definition of Done prüfen
|
|||
|
|
- Abhängigkeiten analysieren
|
|||
|
|
- Risiken prüfen
|
|||
|
|
- Qualitätsgate durchführen
|
|||
|
|
|
|||
|
|
### 23.8 Notification Nodes
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
notification
|
|||
|
|
reminder
|
|||
|
|
escalation
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Zweck:
|
|||
|
|
|
|||
|
|
- informieren
|
|||
|
|
- erinnern
|
|||
|
|
- eskalieren
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 24. Node-Kontrakt
|
|||
|
|
|
|||
|
|
Jeder Node-Typ braucht langfristig einen klaren Vertrag.
|
|||
|
|
|
|||
|
|
Mindestens:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- node_type
|
|||
|
|
- purpose
|
|||
|
|
- allowed_inputs
|
|||
|
|
- outputs
|
|||
|
|
- side_effects
|
|||
|
|
- actor_responsibility
|
|||
|
|
- tenant_context_required
|
|||
|
|
- capability_required
|
|||
|
|
- audit_required
|
|||
|
|
- failure_behavior
|
|||
|
|
- retry_policy optional
|
|||
|
|
- human_approval_required optional
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
AP0.8 muss diesen Kontrakt nicht technisch vollständig bauen.
|
|||
|
|
|
|||
|
|
Aber die Node-Taxonomie muss fachlich dokumentiert sein.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 25. Agentensysteme und KI-Steuerung
|
|||
|
|
|
|||
|
|
### 25.1 Grundentscheidung
|
|||
|
|
|
|||
|
|
Kairo sollte keine vollständige eigene Agentenplattform bauen.
|
|||
|
|
|
|||
|
|
Kairo sollte Agentensysteme als Provider anbinden können.
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Kairo:
|
|||
|
|
- entscheidet Kontext
|
|||
|
|
- wählt Methode/Hook/Workflow
|
|||
|
|
- erzeugt Auftrag
|
|||
|
|
- prüft Rückgabe
|
|||
|
|
- speichert Ergebnis
|
|||
|
|
- erzeugt Signal / Action / Evidence
|
|||
|
|
|
|||
|
|
Agentensystem:
|
|||
|
|
- führt Analyse oder Task aus
|
|||
|
|
- erzeugt Vorschlag oder Ergebnis
|
|||
|
|
- liefert strukturierte Antwort zurück
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 25.2 Agenten sind Actors
|
|||
|
|
|
|||
|
|
Agenten bleiben im Kairo-Modell Actors.
|
|||
|
|
|
|||
|
|
Das ist wichtig für:
|
|||
|
|
|
|||
|
|
- Verantwortlichkeit
|
|||
|
|
- Audit
|
|||
|
|
- Assignments
|
|||
|
|
- Sichtbarkeit
|
|||
|
|
- Tenant-Kontext
|
|||
|
|
|
|||
|
|
### 25.3 Agenten führen Steps aus, aber steuern nicht autonom Kairo
|
|||
|
|
|
|||
|
|
Agenten dürfen nicht ohne Gate:
|
|||
|
|
|
|||
|
|
- Roadmaps verändern
|
|||
|
|
- Actions löschen
|
|||
|
|
- strategische Priorität ändern
|
|||
|
|
- Steering Method ändern
|
|||
|
|
- Workflow Definitions ändern
|
|||
|
|
- Rechte ändern
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 26. AP0.4 Prompt Registry
|
|||
|
|
|
|||
|
|
AP0.4 ist als Prompt-/Workflow-Vorleistung vorhanden.
|
|||
|
|
|
|||
|
|
Im neuen Zielbild wird AP0.4 eingeordnet als:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Prompt Registry
|
|||
|
|
→ später verwendbar für Node-Typ `llm_prompt`
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
AP0.4 soll aktuell nicht weiter als Produktfeature ausgebaut werden.
|
|||
|
|
|
|||
|
|
Es bleibt eingefroren, bis die Method-/Workflow-Schicht fachlich und strukturell steht.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 27. Welche Methoden sollen fachlich möglich sein?
|
|||
|
|
|
|||
|
|
### 27.1 Produktentwicklung
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Domain: product_development
|
|||
|
|
Lifecycle: product/project hybrid
|
|||
|
|
Methoden:
|
|||
|
|
- milestone_driven
|
|||
|
|
- cycle_based
|
|||
|
|
- kanban
|
|||
|
|
- hybrid
|
|||
|
|
Roadmap:
|
|||
|
|
- features
|
|||
|
|
- architecture
|
|||
|
|
- releases
|
|||
|
|
- validation
|
|||
|
|
- risks
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 27.2 Buchentwicklung
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Domain: content_development
|
|||
|
|
Lifecycle: content_development_lifecycle
|
|||
|
|
Methoden:
|
|||
|
|
- chapter_based
|
|||
|
|
- draft_review
|
|||
|
|
- research_driven
|
|||
|
|
Roadmap:
|
|||
|
|
- chapters
|
|||
|
|
- research topics
|
|||
|
|
- drafts
|
|||
|
|
- reviews
|
|||
|
|
- publication steps
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 27.3 Persönliche Reifegradentwicklung
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Domain: maturity_development
|
|||
|
|
Lifecycle: maturity_lifecycle
|
|||
|
|
Methoden:
|
|||
|
|
- maturity_progression
|
|||
|
|
- habit_based
|
|||
|
|
- coaching_cycle
|
|||
|
|
Roadmap:
|
|||
|
|
- maturity stages
|
|||
|
|
- capabilities
|
|||
|
|
- practices
|
|||
|
|
- evidence
|
|||
|
|
- reflection
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 27.4 Programmsteuerung
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Domain: program
|
|||
|
|
Lifecycle: program_lifecycle
|
|||
|
|
Methoden:
|
|||
|
|
- portfolio_coordination
|
|||
|
|
- dependency_management
|
|||
|
|
- benefit_realization
|
|||
|
|
Roadmap:
|
|||
|
|
- strategic outcomes
|
|||
|
|
- initiatives
|
|||
|
|
- dependencies
|
|||
|
|
- resources
|
|||
|
|
- benefits
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 27.5 Routinen/Betrieb
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Domain: routine_operations
|
|||
|
|
Lifecycle: routine_lifecycle
|
|||
|
|
Methoden:
|
|||
|
|
- recurring_control
|
|||
|
|
- checklist_based
|
|||
|
|
- exception_driven
|
|||
|
|
Roadmap:
|
|||
|
|
- routines
|
|||
|
|
- cycles
|
|||
|
|
- checks
|
|||
|
|
- exceptions
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 28. AP0.8 – Konsequenz
|
|||
|
|
|
|||
|
|
Der bestehende AP0.8-Entwurf muss vor Umsetzung überarbeitet werden.
|
|||
|
|
|
|||
|
|
Der bisherige Entwurf mit:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Attention
|
|||
|
|
Blocker
|
|||
|
|
BacklogItem
|
|||
|
|
Milestone
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
ist fachlich nicht falsch, aber zu eng.
|
|||
|
|
|
|||
|
|
Neuer AP0.8-Titel:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
AP0.8 – Steering & Development Roadmap Foundation
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
AP0.8 soll nicht die vollständige Engine bauen, aber die tragfähige Grundlage schaffen.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 29. AP0.8 Soll-Scope
|
|||
|
|
|
|||
|
|
### Teil 0 – Engine Decision dokumentieren
|
|||
|
|
|
|||
|
|
Dieses Dokument oder eine verdichtete Fassung ins Repo übernehmen.
|
|||
|
|
|
|||
|
|
### Teil 1 – Steering Context Foundation
|
|||
|
|
|
|||
|
|
Implementieren oder zumindest als Migration vorbereiten:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
steering_contexts
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Minimal aktiv:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
scope_type = initiative
|
|||
|
|
scope_type = roadmap_item
|
|||
|
|
scope_type = action optional
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### Teil 2 – Roadmap Foundation
|
|||
|
|
|
|||
|
|
Implementieren:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
roadmaps
|
|||
|
|
roadmap_lanes
|
|||
|
|
roadmap_items
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Milestone wird RoadmapItem.
|
|||
|
|
|
|||
|
|
### Teil 3 – BacklogItem
|
|||
|
|
|
|||
|
|
Implementieren:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
backlog_items
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Mit optionalem Bezug:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
roadmap_item_id
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### Teil 4 – Blocker
|
|||
|
|
|
|||
|
|
Implementieren:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
blockers
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Mit optionalem Bezug:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
action_id
|
|||
|
|
roadmap_item_id
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### Teil 5 – Signals
|
|||
|
|
|
|||
|
|
Implementieren:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
AttentionItem als Read Model
|
|||
|
|
NextActionCandidate als Read Model
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Regelbasiert, ohne KI.
|
|||
|
|
|
|||
|
|
### Teil 6 – UI minimal
|
|||
|
|
|
|||
|
|
Im Vorhaben-Detail:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Roadmap
|
|||
|
|
Backlog
|
|||
|
|
Blocker
|
|||
|
|
Attention
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Keine große Designer-UI.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 30. AP0.8 Nicht-Scope
|
|||
|
|
|
|||
|
|
Nicht bauen:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- vollständige Workflow Runtime
|
|||
|
|
- Workflow Designer
|
|||
|
|
- Method Designer
|
|||
|
|
- tenant-konfigurierbare Method Profiles
|
|||
|
|
- vollständige Node Execution Engine
|
|||
|
|
- LLM-Step-Ausführung
|
|||
|
|
- Agentenlaufsteuerung
|
|||
|
|
- MCP
|
|||
|
|
- WBS-Generator
|
|||
|
|
- Dependency Engine
|
|||
|
|
- DoD Engine
|
|||
|
|
- vollständige Evidence-/Review-Engine
|
|||
|
|
- Kanban Board
|
|||
|
|
- Sprint Board
|
|||
|
|
- Burndown
|
|||
|
|
- Velocity
|
|||
|
|
- Gantt
|
|||
|
|
- Kalenderintegration
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 31. AP0.8 Muss aber vorbereiten
|
|||
|
|
|
|||
|
|
AP0.8 muss fachlich und strukturell vorbereiten:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Hook Points
|
|||
|
|
- Signal Types
|
|||
|
|
- RoadmapItem-Typen
|
|||
|
|
- Steering Domains
|
|||
|
|
- Lifecycle Models
|
|||
|
|
- Method Keys
|
|||
|
|
- spätere Workflow Node Types
|
|||
|
|
- spätere Agent Provider
|
|||
|
|
- spätere LLM Prompt Steps
|
|||
|
|
- spätere Tool/MCP Steps
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 32. Komplexitätsregel für Implementierung
|
|||
|
|
|
|||
|
|
Bei jeder Implementierungsentscheidung gilt:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Baue nur, was für AP0.8 sichtbar nutzbar oder strukturell notwendig ist.
|
|||
|
|
Dokumentiere, was später kommt.
|
|||
|
|
Verbaue spätere Workflow-/Method-/Agent-Fähigkeit nicht.
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Das bedeutet:
|
|||
|
|
|
|||
|
|
- Roadmap Foundation ja
|
|||
|
|
- Backlog/Blocker ja
|
|||
|
|
- Attention/NextAction ja
|
|||
|
|
- Workflow Runtime nein
|
|||
|
|
- KI-Ausführung nein
|
|||
|
|
- Method Designer nein
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 33. Offene AP0.8-Entscheidungen
|
|||
|
|
|
|||
|
|
Vor finalem AP0.8-Prompt zu entscheiden:
|
|||
|
|
|
|||
|
|
1. Wird `steering_contexts` in AP0.8 wirklich implementiert oder nur dokumentiert?
|
|||
|
|
2. Wird `actions.roadmap_item_id` direkt ergänzt?
|
|||
|
|
3. Wird `backlog_items.roadmap_item_id` direkt ergänzt?
|
|||
|
|
4. Wird `blockers.roadmap_item_id` direkt ergänzt?
|
|||
|
|
5. Welche `roadmap_item.status`-Werte gelten initial?
|
|||
|
|
6. Welche `roadmap_item.item_type`-Werte gelten initial?
|
|||
|
|
7. Welche `roadmap_type`-Werte gelten initial?
|
|||
|
|
8. Welche minimalen UI-Elemente reichen?
|
|||
|
|
9. Wird AP0.8 in einem großen Auftrag oder in AP0.8a/b/c umgesetzt?
|
|||
|
|
10. Welche Capabilities werden eingeführt?
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 34. Empfohlene Initialwerte
|
|||
|
|
|
|||
|
|
### 34.1 roadmap_item.status
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
planned
|
|||
|
|
active
|
|||
|
|
at_risk
|
|||
|
|
completed
|
|||
|
|
moved
|
|||
|
|
discarded
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 34.2 roadmap_item.item_type
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
milestone
|
|||
|
|
maturity_stage
|
|||
|
|
feature
|
|||
|
|
capability
|
|||
|
|
phase
|
|||
|
|
release
|
|||
|
|
chapter
|
|||
|
|
research_topic
|
|||
|
|
work_cycle
|
|||
|
|
learning_step
|
|||
|
|
review_gate
|
|||
|
|
decision_gate
|
|||
|
|
architecture_step
|
|||
|
|
dependency
|
|||
|
|
risk_reduction
|
|||
|
|
custom
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 34.3 roadmap_type
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
product
|
|||
|
|
program
|
|||
|
|
project
|
|||
|
|
content
|
|||
|
|
personal_development
|
|||
|
|
maturity
|
|||
|
|
operations
|
|||
|
|
learning
|
|||
|
|
custom
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 34.4 steering_domain
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
program
|
|||
|
|
project
|
|||
|
|
product_development
|
|||
|
|
content_development
|
|||
|
|
personal_development
|
|||
|
|
maturity_development
|
|||
|
|
routine_operations
|
|||
|
|
audit_review
|
|||
|
|
agent_execution
|
|||
|
|
custom
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
### 34.5 lifecycle_model
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
project_lifecycle
|
|||
|
|
program_lifecycle
|
|||
|
|
product_development_lifecycle
|
|||
|
|
content_development_lifecycle
|
|||
|
|
maturity_lifecycle
|
|||
|
|
routine_lifecycle
|
|||
|
|
review_lifecycle
|
|||
|
|
agent_run_lifecycle
|
|||
|
|
custom_lifecycle
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 35. Zusammenfassung der Entscheidung
|
|||
|
|
|
|||
|
|
Kairo soll langfristig sehr unterschiedliche Entwicklungen steuern können:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
- Softwareprodukt
|
|||
|
|
- Buch
|
|||
|
|
- persönliche Fähigkeit
|
|||
|
|
- Reifegradziel
|
|||
|
|
- Programm
|
|||
|
|
- Routine
|
|||
|
|
- Review/Audit
|
|||
|
|
- Agentenarbeit
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Dafür reicht weder eine einfache Aufgabenliste noch eine rein projektzentrierte Lifecycle-Logik.
|
|||
|
|
|
|||
|
|
Die zentrale Architekturentscheidung lautet:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Kairo baut eine universelle fachliche Steuerungsengine mit:
|
|||
|
|
- Steerable Objects
|
|||
|
|
- Steering Contexts
|
|||
|
|
- Steering Domains
|
|||
|
|
- Lifecycle Models
|
|||
|
|
- Steering Methods
|
|||
|
|
- Method Profiles
|
|||
|
|
- Development Roadmaps
|
|||
|
|
- Hook Points
|
|||
|
|
- Signals
|
|||
|
|
- später Workflow Nodes
|
|||
|
|
- später Agent-/LLM-/Tool-Steps
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Aber:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
AP0.8 baut nur die erste tragfähige Foundation:
|
|||
|
|
- Steering Context vorbereiten
|
|||
|
|
- Roadmap Foundation
|
|||
|
|
- Backlog
|
|||
|
|
- Blocker
|
|||
|
|
- Attention / NextAction als erste Signals
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Damit bleibt Kairo universell erweiterbar, ohne sofort eine unbeherrschbare Workflow-/Agentenplattform zu werden.
|
|||
|
|
|
|||
|
|
---
|
|||
|
|
|
|||
|
|
## 36. Nächster Schritt
|
|||
|
|
|
|||
|
|
Auf Basis dieses Dokuments sollte der bestehende AP0.8-Entwurf ersetzt oder deutlich überarbeitet werden.
|
|||
|
|
|
|||
|
|
Empfohlenes Folgeartefakt:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
Sprint0_AP0_8_Assignment_v0.2.md
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
Empfohlener Titel:
|
|||
|
|
|
|||
|
|
```text
|
|||
|
|
AP0.8 – Steering & Development Roadmap Foundation
|
|||
|
|
```
|